A method for resolving sources of urban winter atmospheric methane based on carbon isotopes

Abstract

The present application relates to atmospheric environment monitoring and pollution source analysis technical field, disclose a kind of based on carbon isotope's urban winter atmospheric methane source analysis method, comprising the following steps: according to winter energy structure, industry layout, policy and regulation and meteorological condition lock main control artificial emission source, generate and include the mean value and standard deviation of each main control artificial emission source stable carbon isotope characteristic value Localized end member database file;Determination of atmospheric sample, extract methane concentration sequence and stable carbon isotope ratio sequence;Establish linear regression model with the difference between observation concentration and background concentration as independent variable, with the product of stable carbon isotope ratio and observation concentration as dependent variable, obtain the stable carbon isotope characteristic value and standard error of new mixed source;Load bayesian mixed topology model, calculate output relative contribution ratio.The present application eliminates the interference of low temperature non-active source, corrects the fitting deviation caused by hardware measurement, realizes the objective quantitative analysis of the relative contribution ratio of multiple sources.

Description

Technical Field

[0001] This invention relates to the field of atmospheric environmental monitoring and pollution source apportionment technology, specifically to a method for apportioning the sources of methane in urban winter atmosphere based on carbon isotopes. Background Technology

[0002] As a key greenhouse gas affecting global climate change, source apportionment is the foundation for formulating precise emission reduction policies for cities. Source tracing using stable carbon isotope ratios is currently the mainstream method for identifying methane emission sources. By measuring the isotopic characteristic fingerprints of the ambient atmosphere and various pollution sources, a mixing equilibrium model is established to quantitatively assess the contribution ratio of different source types.

[0003] However, existing technical solutions still have limitations in practical applications when conducting winter atmospheric methane source apportionment at the city scale. First, atmospheric methane emissions have obvious seasonal characteristics, and existing end-member library construction methods often lack a deep integration with the target city's winter energy structure and meteorological conditions. They fail to identify and eliminate emission sources such as waste disposal and agricultural activities, which have severely suppressed biological activity under low-temperature conditions. This practice of including inactive sources in the apportionment model increases the computational redundancy of the model and leads to biases in the source identification results.

[0004] In the process of extracting characteristic values ​​of newly added mixed source isotopes using observational data, conventional techniques employ linear regression using the one-way least squares method. However, since online monitoring instruments such as cavity ring-down spectrometers have hardware random errors when measuring gas concentrations and isotope ratios, this approach, which only considers the error of the vertical axis dependent variable while ignoring the error of the horizontal axis independent variable, will cause a systematic shift in the slope and intercept of the fitted line, making it difficult to obtain characteristic values ​​that reflect the true atmospheric conditions.

[0005] When using single isotope data to analyze three or more main emission sources, the analysis system mathematically constitutes an underdetermined equation. Traditional algebraic methods cannot obtain a unique analytical solution. Existing Bayesian models, when dealing with such problems, often rely excessively on artificially set prior empirical distributions and lack scientific constraints on the isotope fractionation effect during winter atmospheric transport. This results in a lack of objectivity in the final output of the contribution rates of each source, making it difficult to meet the requirements of high-precision source analysis. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for apportioning the sources of atmospheric methane in urban winters based on carbon isotopes. This method solves the problems of existing methane source apportionment methods, such as failing to incorporate winter environmental factors to eliminate inactive biological sources, ignoring bidirectional physical measurement errors leading to eigenvalue fitting biases, and the difficulty in objectively quantifying and solving multi-source mixed underdetermined equations.

[0007] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a method for apportioning the sources of atmospheric methane in urban areas during winter based on carbon isotopes. The method includes the following steps: Based on the target city's energy structure, industrial layout, policies and regulations, and meteorological conditions during winter, the main anthropogenic emission sources are identified, and a localized end-member database file containing the mean and standard deviation of stable carbon isotope characteristic values ​​of each main anthropogenic emission source is generated. Measure atmospheric samples and extract methane concentration sequences and stable carbon isotope ratio sequences; The methane concentration from a global atmospheric background reference station is obtained as the background concentration; A linear regression model was established with the difference between the methane concentration sequence and the background concentration as the independent variable and the product of the stable carbon isotope ratio sequence and the methane concentration sequence as the dependent variable, and the characteristic values ​​and standard errors of the newly added mixed source stable carbon isotopes were decoupled. The stable carbon isotope eigenvalues ​​and standard errors of the newly added mixed sources, along with the localized end-member database file, are loaded into the Bayesian mixed topology model to calculate and output the relative contribution ratio of each controlling human to the emission sources.

[0008] Furthermore, the step of identifying the primary anthropogenic emission sources based on the target city's energy structure, industrial layout, policies and regulations, and meteorological conditions during winter includes: obtaining the baseline emission flux of potential methane emission sources in the target city; when the contribution rate of a potential methane emission source to the baseline emission flux in the target season is lower than a preset contribution rate threshold, or when the biodegradation mechanism is significantly inhibited due to objective environmental factors, the potential source is eliminated; identifying refrigeration, heating, or industrial coal mining sources, natural gas system leaks and combustion sources, and coal chemical and industrial emission sources as the primary anthropogenic emission sources; the preset contribution rate threshold ranges from 1% to 5%.

[0009] Before inputting the model, emission sources with low basic contribution rates or those that have lost their activity due to low temperatures are pre-removed, thereby reducing the computational dimensionality of the subsequent probabilistic model and minimizing interference.

[0010] Furthermore, the step of eliminating the potential source when the biodegradation mechanism is significantly inhibited due to objective environmental factors includes: when the obtained winter temperature is lower than a preset temperature threshold, it is determined that the biological activity of methanogens in wastewater treatment is significantly inhibited, thereby eliminating the waste treatment source, wherein the temperature threshold ranges from 0°C to 4°C; based on the seasonal characteristics of crop fallow, wetland freezing, and vegetation dormancy, it is determined that the emissions from agricultural activity sources, wetland sources, and vegetation sources are significantly inhibited and their basic emission contribution rate is lower than the preset contribution rate threshold, and pruning and elimination are performed.

[0011] By introducing temperature and seasonal characteristic parameters, the elimination conditions for sources such as agriculture, wetlands, vegetation, and sewage treatment are clarified, thereby improving the accuracy of source screening.

[0012] Furthermore, the step of generating a localized end-member database file containing the mean and standard deviation of stable carbon isotope characteristic values ​​of each main-controlling emission source includes: assuming that the stable carbon isotope ratio variable of any of the main-controlling emission sources follows a normal distribution; when extracting measurement values ​​from historical literature data, if the obtained data sample size is lower than a preset sample size filtering limit, then controlling the on-site sampling device to collect gas samples for measurement; the preset sample size filtering limit is not less than 15.

[0013] By combining literature data with field measurement data, the problem of insufficient sample size from a single data source is overcome, providing reliable end-metadata for the analytical model.

[0014] Furthermore, before the step of measuring the atmospheric sample, the constraints for obtaining the atmospheric sample include: setting the air inlet height above the ground to 15 to 20 meters; and setting the daily physical sampling time window to the interval between 13:00 and 15:00 in the afternoon.

[0015] Sampling was scheduled for the afternoon, taking advantage of the thorough convective mixing in the atmospheric boundary layer during this period. This ensured that the samples obtained accurately reflected the average concentration at the city scale, avoiding the interference of local emissions accumulation caused by temperature inversions at night or in the early morning.

[0016] Furthermore, the atmospheric sample measurement steps are performed using a cavity ring-down spectrometer. The measurement steps include: controlling the continuous inlet gas flow rate of the cavity ring-down spectrometer to be 20 mL / min to 30 mL / min; discarding the initial measurement data and extracting only the data from the stable period to calculate the average methane concentration and the average stable carbon isotope ratio of the environmental atmospheric sample, and integrating them into the methane concentration sequence and stable carbon isotope ratio sequence after the observation period ends; the initial measurement data discarding time is set to 5 to 10 minutes.

[0017] By eliminating initial fluctuation data and removing interference from residual gas in the pipeline and the pressure build-up process in the resonant cavity on the measurement baseline, the stability of the measurement data is ensured.

[0018] Furthermore, the step of decoupling the newly added mixed-source stable carbon isotope eigenvalues ​​and standard errors includes: using a bivariate bidirectional error regression algorithm to fit the linear regression model to a straight line; introducing hardware measurement errors in the horizontal and vertical directions during the iterative calculation process; calculating the measurement error weighting factors of each observation data point in the two axes respectively; constructing a comprehensive equivalent weighting coefficient containing bidirectional errors; and decoupling the newly added mixed-source stable carbon isotope eigenvalues ​​based on the slope term of the fitted straight line calculated by the optimal fitting.

[0019] Introducing bidirectional measurement errors into regression calculations corrects the fitting bias caused by neglecting independent variable errors in conventional least squares methods, thereby improving the accuracy of eigenvalue decoupling.

[0020] Furthermore, after the step of decoupling the newly added mixed-source stable carbon isotope eigenvalues ​​and standard errors, the method further includes a step of performing regression quality filtering. The newly added mixed-source stable carbon isotope eigenvalues ​​are deemed valid and retained as output only if the following three-level constraint indices are met simultaneously: First-level index: the difference between the highest and lowest methane concentrations in the methane concentration sequence during the observation period is greater than the concentration range threshold of 0.15 ppm to 0.35 ppm; Second-level index: the coefficient of determination of the optimized fitted line is greater than the coefficient of determination threshold of 0.80 to 0.90; Third-level index: the standard error of the output slope is less than the upper limit threshold of the standard error of 1.5‰ to 2.5‰.

[0021] By using three levels of index constraints, invalid regression results caused by atmospheric static accumulation or instrument noise are eliminated, ensuring the reliability of the feature values ​​input into subsequent models.

[0022] Furthermore, the step of loading the newly added mixed source stable carbon isotope eigenvalues, standard errors, and the localized end-member database file into the Bayesian mixed topology model includes: constructing a probabilistic mass balance control equation, and setting the isotope fractionation coefficients and their standard deviations of each of the main anthropogenic emission sources during atmospheric transport to zero; constructing an informationless uniform prior distribution model using a Dirichlet distribution, wherein the values ​​of the concentration hyperparameters corresponding to each of the main anthropogenic emission sources in the Dirichlet distribution are all set to 1.

[0023] Based on the low temperature and weak light conditions in winter, the fractionation effect is ignored, and an information-free prior distribution is used to assign equal initial weights to each source, so as to avoid the bias interference of human experience on the model calculation.

[0024] Furthermore, the step of calculating and outputting the relative contribution ratio of each controlling agent to the emission source includes: running a Markov chain Monte Carlo iterative algorithm program to randomly sample in the parameter space, starting at least three independent Markov chains in parallel, and calculating the potential scaling factor of each parameter in real time; when the potential scaling factor meets the convergence judgment threshold condition, the judgment model reaches a stable distribution, performing mathematical statistics on the retained effective sample chains, and outputting the posterior probability mean of the relative contribution ratio of each controlling agent to the emission source.

[0025] The Markov chain Monte Carlo algorithm is used to process the mathematical underdetermined equations composed of single isotope data and multiple master control sources, so as to achieve quantitative analysis of the relative contribution ratio of multiple sources.

[0026] This invention provides a method for apportioning the sources of atmospheric methane in urban winters based on carbon isotopes. It has the following beneficial effects: 1. This invention combines the meteorological conditions and energy structure of the target city in winter to pre-screen potential emission sources. By setting temperature thresholds and contribution rate thresholds, inactive sources such as waste treatment sources and agricultural sources whose biological activity is significantly suppressed are removed from the calculation list. This step removes the interference of non-controlling sources on model analysis, reduces the parameter calculation space of the multi-source analysis model, and improves the accuracy of locking the main anthropogenic sources of atmospheric methane in winter.

[0027] 2. When decoupling the stable carbon isotope eigenvalues ​​of newly added mixed sources, this invention employs a bivariate bidirectional error regression algorithm. It simultaneously introduces two axial hardware measurement errors—the difference between methane concentration and background concentration, and the product of stable carbon isotope ratio and methane concentration—for weight allocation. Combined with a set three-level regression quality constraint index, it performs data filtering, corrects the fitting bias caused by neglecting independent variable errors in the traditional one-way least squares method, eliminates invalid data caused by instrument noise, and improves the numerical accuracy of eigenvalue extraction.

[0028] 3. This invention constructs a Bayesian hybrid topology model and uses the Markov chain Monte Carlo algorithm for iterative sampling. It combines an informationless uniform prior distribution model to assign equal initial weights to each major human-controlled emission source. In conjunction with the characteristics of winter environment, the isotope fractionation coefficient is set to zero to avoid the bias interference of human experience. This solves the problem of solving mathematical underdetermined equations when a single carbon isotope tracer faces more than three major emission sources, and achieves objective quantitative analysis of the relative contribution ratio of multiple sources. Attached Figure Description

[0029] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a schematic diagram of the method flow of the present invention; Figure 3 This is a schematic diagram of the process for quantitatively estimating the contribution rate of pollution sources according to the present invention; Figure 4 This is a schematic diagram of the process for calculating the eigenvalues ​​of mixed sources according to the present invention; Figure 5 This is a curve showing the extraction of new hybrid source features based on bidirectional error-weighted regression according to the present invention.

[0030] Among them, 110 is the end-member library construction module; 120 is the sample collection and measurement module; 130 is the feature value extraction module; and 140 is the contribution rate estimation module. Detailed Implementation

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] See attached document Figure 1 This invention provides a carbon isotope-based system for apportioning the source of atmospheric methane in urban winters. The system includes an end-member library construction module 110, a sample collection and measurement module 120, a feature value extraction module 130, and a contribution rate estimation module 140.

[0033] The sample acquisition and measurement module 120 is equipped with physical hardware, which includes an ambient air intake device deployed on the top of a representative building in the city center, a 5-liter aluminum foil gas sampling bag, a 12V small diaphragm gas pump, a high-purity nitrogen cleaning gas path assembly, an optical cavity ring-down spectrometer, and a matching standard gas calibration high-pressure cylinder. The physical hardware is spatially limited to a position 15 to 20 meters above the ground and avoiding local plumes. The physical hardware is responsible for the quantitative extraction of ambient atmosphere, airtight storage, and online physical measurement of methane concentration and stable carbon isotope ratio at the physical level, converting the gaseous material flow into digital electrical signals characterizing physical properties.

[0034] The end-member library construction module 110, the feature value extraction module 130, and the contribution rate estimation module 140 are all implemented based on the physical computing hardware unit. The physical computing hardware unit includes a central processing unit, a high-frequency random access memory, a large-capacity non-volatile data memory, an input / output interface, an external network communication interface or a human-computer interaction input module, and an internal data transmission bus. The physical computing hardware unit communicates with the data output terminal of the optical cavity ring-down spectrometer through the input / output interface to receive the digital measurement sequence in real time. Simultaneously, it accesses the urban environmental database or receives historical meteorological, policy, regulatory, and inventory data imported by operators through an external network communication interface or human-computer interaction input module; and calls the mathematical statistical control program and probabilistic inversion algorithm stored in the data storage in the central processing unit. This hardware architecture provides a physical carrier for data acquisition, signal transmission, logical operation, and statistical result storage for a carbon isotope-based method for apportioning urban winter atmospheric methane sources.

[0035] See attached document Figure 2 This invention provides a method for apportioning the source of atmospheric methane in urban winters based on carbon isotopes, comprising the following steps: S1, construct a localized end-element library. The end-element library construction module 110 executes boundary control logic. Based on the target city's energy structure, industrial layout, policies and regulations, and meteorological conditions in winter, it removes sources with contribution rates below a preset threshold or those in a state of biological activity inhibition in winter. It locks in sources of cooling, heating, or industrial coal mining, natural gas system leaks and combustion, and coal chemical and industrial emissions. Through historical literature data extraction and on-site sampling and measurement of typical points, it generates a localized end-element database file containing the mean and standard deviation of stable carbon isotope characteristic values ​​of each main-controlling human emission source in a non-volatile data storage.

[0036] S2, synchronously collects and measures atmospheric samples. Within a preset continuous observation period, the sample collection and measurement module 120 drives a 12V small diaphragm gas pump to fill a 5-liter aluminum foil gas sampling bag with ambient air during a specific time window. The gas in the bag is then introduced into the physical optical resonant cavity of the optical cavity ring-down spectrometer for polarization laser measurement. The photoelectric sensor inside the spectrometer converts the ring-down light signal into a digital signal. After removing the initial residual data from the pipeline, the module extracts the methane concentration sequence and stable carbon isotope ratio sequence reflecting the actual atmosphere. After multiple cycles of sampling and measurement, the module obtains a set of effective observation sequences containing multiple discrete samples and sends them to the computing hardware via the data transmission bus.

[0037] S3, calculate the characteristic value of the mixed source. The characteristic value extraction module 130 reads the concentration sequence and isotope ratio sequence through the internal data transmission bus. The central processing unit calls the bivariate bidirectional error regression algorithm program to establish a linear regression model with the difference between methane concentration and background concentration as the independent variable and the product of stable carbon isotope ratio and methane concentration sequence as the dependent variable. The characteristic value of stable carbon isotope of the newly added mixed source after eliminating atmospheric background interference is decoupled by the slope term of the fitted line, and the corresponding standard error is output synchronously.

[0038] S4, quantitatively estimate the contribution rate of pollution sources. The contribution rate estimation module 140 loads the newly added stable carbon isotope eigenvalues, standard errors, and pre-stored localized end-member database files of mixed sources into the Bayesian mixed topology model. The central processing unit runs the Markov chain Monte Carlo iterative algorithm program, performs high-dimensional probability distribution integration under the boundary conditions of uninformed uniform prior distribution and zero fractionation coefficient, calculates and outputs the posterior mean and confidence interval data of the relative contribution ratio of each main anthropogenic emission source, and completes the quantitative analysis of the source of atmospheric methane in the city during winter.

[0039] In specific implementations, step S1 provided by this invention may include the following sub-steps: S11, perform boundary screening and prior pruning of potential sources. The end-member library construction module 110 retrieves the target city's energy structure data, industrial layout data, policy and regulatory texts, and meteorological data in winter. According to the preset screening criteria, when the baseline emission flux contribution rate of a certain potential methane emission source in the target season is lower than the preset contribution rate threshold, or when the biodegradation mechanism is significantly inhibited due to objective environmental factors, the end-member library construction module 110 cuts off the connection of the potential source node in the data network to achieve prior pruning and isolation. In this embodiment, the preset contribution rate threshold ranges from 1% to 5%, preferably 2%. The contribution rate threshold is determined by statistical calculation based on the baseline emission flux ratio of each source type in the target city's winter air pollutant source apportionment inventory data of the previous year.

[0040] Specifically, for waste treatment sources, based on the target city's implemented policy of complete incineration and zero landfill of municipal solid waste, it is determined that the landfill gas emissions have been reduced to a level that is negligible in the environment. At the same time, winter temperature data is retrieved. When the temperature is lower than a preset temperature threshold, it is determined that the biological activity of methanogenic bacteria in wastewater treatment is significantly inhibited, thereby eliminating the waste treatment source. In this embodiment, the preset temperature threshold ranges from 0°C to 4°C, preferably 0°C.

[0041] For agricultural activity sources, wetland sources, and vegetation sources, winter natural phenological data are retrieved. Based on the seasonal characteristics of crop fallow, wetland freezing, and vegetation dormancy, it is determined that methane emissions are significantly suppressed and their baseline emission contribution rate is lower than the preset contribution rate threshold. Pruning and removal are then performed. For biomass combustion sources, local winter straw burning control regulations are retrieved. Combined with the spatial characteristics of the urban center area, it is determined that the emission proportion is lower than the contribution rate threshold. Pruning and removal are then performed. Through the aforementioned screening and filtering, the end-element library construction module 110 locks refrigeration, heating, heating or industrial coal mining sources, natural gas system leakage and combustion sources, and coal chemical and industrial emission sources as the three main controlled anthropogenic emission sources in the target area during winter.

[0042] S12 defines the industry and physical boundaries of the main anthropogenic emission sources. The end-element library construction module 110 accurately defines the spatial range and process chain of the three types of main anthropogenic emission sources. The boundary of the coal mining source is limited to the coal production process, up to the point where raw coal or washed coal leaves the factory. The physical entities include mine ventilation openings, gas drainage stations and mining face vents, which are used to capture the original coalbed methane signals of thermal or mixed origin.

[0043] The boundary between natural gas system leaks and combustion sources is defined in the transmission, distribution, and end-use stages of natural gas as a primary energy source. Physical entities include city natural gas gate stations, high and low pressure regulating stations, residential and commercial gas pipeline interfaces, and natural gas combustion equipment, used to capture methane signals from specific gas sources during transportation and incomplete combustion. The boundary between coal chemical and industrial emission sources is defined in industrial production processes that use coal or natural gas as raw materials for chemical conversion or as direct fuel. Physical entities include coking furnace tail gas emission outlets, blast furnace interfaces in steel smelting, and process exhaust valves in chemical synthesis towers. The process cutoff boundary is defined as starting from coarse coal crushing and depolymerization and ending with multi-stage condensation separation or tail gas venting, used to capture methane signals in high-temperature conversion or complex industrial waste gases.

[0044] S13, Capture and integrate localized end-member feature fingerprint data. The end-member library construction module 110, for the three types of main controlled anthropogenic emission sources selected, collects stable carbon isotope ratios by combining historical literature meta-analysis with on-site sampling and measurement of typical points. For any main controlled anthropogenic emission source, its isotope characteristic distribution is considered to follow a normal distribution in mathematical statistics. The mathematical control equation is as follows: ; In the formula, For the first The stable carbon isotope ratio variable of the main controller of the emission source; This is the identifier for the normal distribution function in probability and statistics. For the first The average stable carbon isotope ratio of each major human emission source; For the first Standard deviation of stable carbon isotope ratios for each major source of emission, expressed in ‰; The index number is the one used to identify the main source of emissions.

[0045] During the data acquisition process, for coal mining sources, the end-member library construction module 110 retrieves existing measured literature on the coalfield where the target city is located, sets a sample size filter limit, and the range of the sample size filter limit is no less than 15, preferably no less than 30, and extracts the measured values ​​of stable carbon isotope ratios. If the literature data is lower than the preset sample size filter limit, the on-site sampling device is controlled to collect gas samples at representative mine ventilation openings, and the measurements are performed according to existing gas chromatography-isotope ratio mass spectrometry analysis methods. For natural gas system leakage and combustion sources, physical gas bag sampling points are established at the city natural gas gate station and residential area pipeline interfaces. Typical escaping samples are physically collected using gas bags, and their stable carbon isotope ratios are measured using a cavity ring-down spectrometer. For coal chemical and industrial emission sources, sampling pipelines are laid at the process tail gas emission outlets of coking and steel enterprises to collect tail gas samples and measure them. Before each physical measurement, the measuring instruments are calibrated using standard gases that have been traced through international standard reference materials.

[0046] The end-member library construction module 110 summarizes and calculates the acquired multi-source discrete data, calculates the arithmetic mean of each source as the mean, calculates the sample standard deviation as the standard deviation, and records the corresponding total sample size. Finally, the end-member library construction module 110 organizes these parameters into a structured standard table containing source name, stable carbon isotope mean, stable carbon isotope standard deviation, sample size, and data source type, and generates the corresponding localized end-member database file in a non-volatile data storage, completing the construction of the entire end-member library and providing a digital benchmark for subsequent mixed source feature value extraction and contribution rate estimation.

[0047] In specific implementations, step S2 provided by the present invention may include the following sub-steps: S21, determine the spatial topology and environmental constraints of the sampling site. The sample collection and measurement module 120 deploys the air intake physical hardware on the roof of a representative building in the central area of ​​the target city. Through physical decoupling of spatial height and surrounding environment, the collected gas sample is a sample after being fully mixed with the urban atmospheric boundary layer. In this embodiment, the air intake height above the ground ranges from 15 meters to 20 meters. This air intake height is determined based on the average height of buildings in the central area of ​​the target city and the roughness of the urban underlying surface. Its increment relative to the average height of surrounding buildings ranges from 2 meters to 5 meters, so that the air intake is above the urban canopy.

[0048] The location of the air inlet meets the physical conditions of being open on all sides, avoiding the direct gas plume impact from local chimney exhaust outlets, building ventilation valves, and the sides of main traffic arteries. Through physical decoupling of spatial height from the surrounding environment, the collected gas samples are samples that have been fully mixed in the urban atmospheric boundary layer, which are used to characterize the overall average signal at the urban regional scale and eliminate the single-point interference of local strong emission sources at the lower level on the macroscopic analysis results.

[0049] S22, Determine the physical and dynamic boundary of the sampling time window. The sample collection and measurement module 120 sets the daily physical sampling time window in the afternoon between 13:00 and 15:00, preferably 14:00. The time window is determined based on the daily variation dynamic curve of the local winter convective boundary layer height. The period of strong convection when the boundary layer height reaches its maximum value and remains stable is selected. The selection of this time window is based on the principle of atmospheric boundary layer dynamics.

[0050] During the winter diurnal cycle, the atmospheric planetary boundary layer reaches its deepest point of the day, and the vertical convection height of the atmosphere reaches its maximum. Vertical convection mixing occurs between the near-surface layer and the upper free atmosphere, and the entire atmosphere is in a uniformly mixed state. Selecting this period for sample collection can dilute and eliminate the artificially high concentration phenomenon caused by the accumulation of temperature inversion layer at night or in the early morning due to weak local surface fugitive sources, ensuring that the acquired data can truly reflect the background and newly added mixing signals at the urban scale.

[0051] S23, Perform physical collection and airtightness control of gas samples. For the physical collection and preservation of gas samples, the existing gas extraction method is used. Before sampling, the sample collection and measurement module 120 uses high-purity nitrogen with a purity greater than or equal to 99.999% to repeatedly rinse and evacuate to a vacuum state an aluminum foil gas sampling bag with a volume of 3 to 10 liters, preferably 5 liters. The number of rinsing times is limited to 3 to 5 times to remove historical residual gas.

[0052] Within the preset sampling time window, start the diaphragm air pump with a pumping rate of 1 to 3 liters per minute. First, use ambient air to physically flush the air inlet pipeline for 2 to 3 minutes. Then, connect the pipeline to the air inlet of the aluminum foil gas sampling bag and stop pumping when the bag is filled to 80% to 90% of its total capacity. Immediately implement physical sealing. While pumping air, use the matching environmental sensors to simultaneously measure and record meteorological data such as ambient temperature, ambient air pressure, real-time wind speed, and real-time wind direction.

[0053] S24, implement online physical measurement of concentration and isotopes based on cavity ring-down spectroscopy. Cavity ring-down spectroscopy technology uses polarized laser of a specific wavelength injected into an optical resonant cavity composed of high reflectivity lenses. By detecting the difference in light intensity decay time in the cavity when there is no gas to be measured, the concentration of methane molecules of different mass numbers is simultaneously and quantitatively analyzed according to the principle of absorption spectroscopy. The sample acquisition and measurement module 120 connects the physically collected aluminum foil gas sampling bag to the air inlet of the cavity ring-down spectrometer. The continuous air intake flow rate is controlled by the built-in flow control valve of the instrument, and the total physical measurement time of a single gas bag sample is set. In order to eliminate the residual gas memory effect inside the air intake pipeline and the hardware instability of the optical resonant cavity in the early stage of air intake, the data processing control program discards the initial measurement data and only extracts the data of the stable period for mean calculation.

[0054] In this embodiment, the continuous inlet gas flow rate ranges from 20 ml / min to 30 ml / min, preferably 25 ml / min; the total physical measurement time for a single gas bag sample ranges from 10 minutes to 15 minutes; the time for capturing stable period data is 5 minutes; the initial measurement data rejection time is determined based on the ratio of the internal volume of the optical resonant cavity of the optical cavity ring-down spectrometer to the continuous inlet gas flow rate, ensuring that the rejection time is greater than the time required for the gas inside the optical resonant cavity to be completely replaced 3 to 5 times, and the rejection time ranges from 5 minutes to 10 minutes.

[0055] By averaging the discrete data points within this stable period, the average methane concentration and the average stable carbon isotope ratio of the environmental atmospheric samples were extracted. The relevant calculation equations are as follows: ; In the formula: This represents the average concentration of methane in actual environmental observation samples. This represents the total number of valid data collection points within a stable period of a single physical sampling. This refers to the index number of the data acquisition point within the stable period of a single physical sampling; For the first time during the stable period Real-time methane concentration measurements at each data collection point.

[0056] ; In the formula: The average stable carbon isotope ratio of methane in actual environmental observation samples; This represents the total number of valid data collection points within a stable period of a single physical sampling. This refers to the index number of the data acquisition point within the stable period of a single physical sampling; For the first time during the stable period Real-time measured values ​​of stable carbon isotope ratios at each data collection point.

[0057] Before each batch of physical measurements, the sample acquisition and measurement module 120 uses standard gas from a high-pressure cylinder to perform multi-point linear calibration on the optical cavity ring-down spectrometer. The standard gas is traced back to the World Meteorological Organization scale to ensure the accuracy of the physical measurement sequence. Finally, the extracted average concentration and average isotope ratio are converted into digital electrical signals, which are then integrated into a dataset containing... The sequence of observation sets of discrete data pairs is sent to the subsequent data processing unit for linear fitting via the internal data transmission bus.

[0058] See attached document Figure 3 In specific implementations, step S3 provided by the present invention may include the following sub-steps: S31, a binary linear mixing control equation based on isotope mass balance is constructed. The feature extraction module 130 receives digital signals such as the average methane concentration and the average stable carbon isotope ratio of the environmental atmospheric sample transmitted from the sample acquisition and measurement module 120. For methane in the atmospheric environment, this embodiment considers it as a two-terminal mixing system composed of background methane and locally newly emitted methane. Based on the principle of mass and isotope conservation, the following linear relationship can be derived: the difference between the observed methane concentration and the background concentration (C obs C bg ) is the independent variable, and the product of the observed stable carbon isotope ratio and the observed concentration is used as the independent variable. Using as the dependent variable, a linear regression equation is constructed, with the specific control equation as follows: ; In the formula: The average stable carbon isotope ratio of methane in actual environmental observation samples; This represents the average concentration of methane in actual environmental observation samples. The methane concentration constant in the regional background atmosphere; The methane stable carbon isotope ratio constant of the regional background atmosphere; The comprehensive stable carbon isotope characteristic value of newly added mixed sources in the city, i.e. the slope of the linear fit.

[0059] S32, a two-way error regression model is used for linear fitting. The feature extraction module 130 uses a bivariate two-way error regression algorithm to perform linear fitting on the aforementioned control equation. Since in actual observation, both isotope ratios and gas concentrations have random measurement errors introduced by the analytical instruments, if the conventional one-way least squares method is used, the error of the horizontal axis independent variable will be ignored, which will cause a systematic bias in the slope estimation. In this embodiment, during the iterative process of regression fitting, hardware measurement errors in the horizontal and vertical axes are introduced simultaneously, and the error weight factors of each observation data point in the two axes are calculated respectively to construct a comprehensive weight function that includes two-way errors.

[0060] To avoid the algorithm getting stuck in a loop dead zone, this embodiment uses the slope of the straight line calculated by ordinary least squares as the initial iteration slope of the bivariate bidirectional error regression algorithm. Since this embodiment uses the Miller-Tans method to construct the linear relationship, its ordinate dependent variable is composed of the product of concentration and isotope ratio, requiring a redefinition of the biaxial measurement standard deviation based on the error propagation principle. Subsequently, based on the optimization objective function and iterative control equation, the optimal fitting slope is calculated using a cyclic iterative approximation method, with a convergence accuracy threshold and a maximum iteration limit pre-set in the iterative control program. In this embodiment, the preset convergence discrimination accuracy threshold ranges from 10. -6 Up to 10 -8 Preferably 10 -7 The determination is based on the matching relationship between the effective number of bits in the computer's double-precision floating-point arithmetic and the measurement resolution of the analytical instrument; the preset maximum number of iterations is limited to 100 to 500 times, preferably 200 times, to achieve a balance between system computing power and fitting convergence. The relevant optimization objective function and iterative control equations are as follows: ; In the formula: This represents the weighted sum of squared residuals objective function value; The total number of valid observation samples participating in the regression calculation; Index numbering for valid observation samples participating in regression calculations; For the first The measurement error weighting factor of each observed sample on the horizontal axis; For the first The adjusted x-coordinate values ​​of the fitted points corresponding to each observed sample; For the first The adjusted x-coordinate values ​​of the fitted points corresponding to each observed sample; For the first Measurement error weighting factor for each observed sample on the vertical axis; For the first The product of the stable carbon isotope ratio of methane and the methane concentration for each observed sample; For the first The adjusted ordinate values ​​of the fitted points corresponding to each observed sample.

[0061] ; In the formula: For the first The measurement error weighting factor of each observed sample on the horizontal axis; Index numbering for valid observation samples participating in regression calculations; For the first The standard deviation of the abscissa value of each observed sample is the hardware standard deviation of the methane concentration measured by the cavity ring-down spectrometer.

[0062] ; In the formula: For the first Measurement error weighting factor for each observed sample on the vertical axis; Index numbering for valid observation samples participating in regression calculations; For the first The equivalent measurement standard deviation of the ordinate values ​​of each observed sample. Since the ordinate is the product of the isotope ratio and concentration, the calculation logic is derived from the error propagation formula: ; In the formula: For the first The equivalent measurement standard deviation of the ordinate values ​​of each observed sample; and These represent the stable carbon isotope ratios and concentration observations, respectively. and These are the standard deviations of the corresponding instrument measurements.

[0063] ; In the formula: For the first Each observation sample includes a comprehensive equivalent weighting coefficient for two-way errors; Index numbering for valid observation samples participating in regression calculations; For the first The measurement error weighting factor of each observed sample on the horizontal axis; For the first Measurement error weighting factor for each observed sample on the vertical axis; This is the slope term of the fitted line updated in each iteration.

[0064] S33, decouple the newly added mixed source stable carbon isotope eigenvalues ​​and filter them using regression quality indicators. The eigenvalue extraction module 130, based on the optimal fitting result after iterative convergence, decouples the slope term of the fitted line. Based on the physical meaning of the Miller-Tans method, this slope term is the comprehensive stable carbon isotope eigenvalue of the newly added local mixed source after excluding background atmospheric interference. The program simultaneously calculates the standard error of this slope term based on the error propagation matrix.

[0065] To ensure that the extracted feature values ​​have the reliability required to support subsequent source apportionment inversion, the feature value extraction module 130 sets a regression quality filtering control logic to eliminate abnormal calculation results caused by excessively stable meteorological conditions during the observation period, resulting in too small concentration fluctuations or severe interference from instrument hardware noise. The regression quality filtering control logic sets three levels of constraint indicators.

[0066] The first level requires that the difference between the highest and lowest methane concentrations during the observation period be greater than a preset concentration variation threshold. This threshold is set based on the assumption that the signal change must be 5 to 10 times greater than the baseline measurement noise of the spectrometer to ensure the signal-to-noise ratio. The second level requires that the coefficient of determination of the fitted straight line be greater than a preset coefficient of determination threshold to ensure the statistical significance of the linear correlation between the two variables. The third level requires that the standard error of the output slope be less than a preset upper limit threshold for standard error to limit the uncertainty tolerance boundary of the feature values. In this embodiment, the concentration variation threshold ranges from 0.15 ppm to 0.35 ppm, preferably 0.20 ppm; the determination coefficient threshold ranges from 0.80 to 0.90, preferably 0.85; and the upper limit of the standard error threshold ranges from 1.5‰ to 2.5‰, preferably 2.0‰. Only when a set of observation calculation results simultaneously meets the above three-level constraint indicators is the result deemed valid and the output retained. The decoupling of the eigenvalues ​​and the calculation equation for their standard error are as follows: ; In the formula: The comprehensive stable carbon isotope characteristic values ​​of newly added mixed sources in the city; This is the slope term of the optimal fitted line output by the bivariate bidirectional error regression algorithm after reaching the convergence discrimination accuracy threshold.

[0067] ; In the formula: The standard error of the newly added integrated stable carbon isotope characteristic values ​​is the slope standard error of the fitted straight line. The total number of valid observation samples participating in the regression calculation; Index numbering for valid observation samples participating in regression calculations; For the first Each observation sample includes a comprehensive equivalent weighting coefficient for two-way errors; The weighted average center value of all valid observation samples along the horizontal axis; For the first The methane concentration value of each observed sample or the difference between the methane concentration and the background concentration.

[0068] The feature extraction module 130 outputs the effective new mixed source stable carbon isotope feature values ​​and their standard errors filtered by the above quality control, and sends them as physical parameters to subsequent modules through the internal data bus for quantitative estimation of source contribution rate.

[0069] See attached document Figure 4 In specific implementations, step S4 provided by the present invention may include the following sub-steps: S41, construct the basic equations of the linear mixture model based on the Bayesian probability framework. The contribution rate estimation module 140 reads the newly added stable carbon isotope eigenvalues ​​and standard errors output by the eigenvalue extraction module 130 through the internal data bus as the target observation input of the system. At the same time, the contribution rate estimation module 140 calls the localized end-member database file generated by the end-member library construction module 110 to extract the average value and standard deviation of the stable carbon isotopes of the three main controlled anthropogenic emission sources as the source fingerprint input of the system.

[0070] For multi-source hybrid systems, in addition to the instrument measurement errors inherent in the target observation values, the characteristic fingerprints of each main control source in the end-member library are not absolutely fixed single values, but have a certain natural fluctuation range. Traditional least squares methods are difficult to handle two-way uncertainties simultaneously. This embodiment adopts a Bayesian probabilistic framework. Based on the law of conservation of isotope mass, the characteristic values ​​of newly added hybrid source isotopes in the atmosphere are represented as a linear weighted sum of the products of the average characteristic values ​​of each main control source isotope and its corresponding emission contribution rate. At the same time, based on the error propagation principle, the variances of each main control source are weighted and summed according to the square of its contribution ratio, thereby propagating the source-end fluctuation uncertainty to the final observation prediction variance. The contribution rate estimation module 140 places this linear hybrid logic into the Bayesian probabilistic framework to construct a probabilistic mass balance control equation. The specific probabilistic observation model is as follows: ; In the formula: The comprehensive stable carbon isotope characteristic values ​​of newly added mixed sources in the city; This is the identifier for the normal distribution function in probability and statistics. The total number of primary human-controlled emission sources; The index number of the emission source controlled by the main person; For the first The relative contribution of each primary source of emissions to the increase of methane in the atmosphere; For the first The average stable carbon isotope ratio of each major human emission source; For the first The primary controller is the isotope fractionation coefficient of the emission source during atmospheric transport; For the first Standard deviation of stable carbon isotope ratios for each major human-controlled emission source.

[0071] S42, Limiting System Boundary Conditions: To ensure the solvability of the Bayesian model and to fit the actual physical environment of the target city, the contribution rate estimation module 140 limits the boundary conditions of the model parameters before performing high-dimensional integration. Regarding the isotope fractionation coefficient, based on the meteorological and chemical characteristics of the target city in winter (low solar radiation intensity and low concentration of oxidants such as hydroxyl radicals in the atmosphere throughout the year), and considering the short physical transmission distance from pollution sources to observation points at the city scale, it is determined that the photochemical degradation of methane molecules during transmission is weak, and the kinetic isotope fractionation effect is negligible. Therefore, the contribution rate estimation module 140 sets the isotope fractionation coefficients and their standard deviations of all major anthropogenic emission sources to zero.

[0072] To ensure system objectivity and avoid introducing empirical biases towards specific pollution sources, the contribution rate estimation module 140 introduces a Dirichlet distribution to construct an informationless uniform prior distribution model. This model ensures that the prior contribution probabilities of the three main controlled human emission sources are in a state of completely equal weight before the system's observation data is input, thus representing the state of no prior bias. In this embodiment, the system limits the value of each concentration hyperparameter in the Dirichlet distribution to 1. The corresponding mathematical expression of the prior distribution is as follows: ; In the formula: The parameter vector representing the relative contribution ratio of the main controlling human emission sources; The total number of primary human-controlled emission sources; This is the identifier for the Dirichlet distribution function in probability and statistics. This represents the concentration hyperparameters corresponding to each major anthropogenic emission source in the Dirichlet distribution.

[0073] S43, perform iteration and post-convergence diagnosis. The contribution rate estimation module 140, based on Bayes' theorem, multiplies the likelihood function of the observed data with the set uninformed uniform prior distribution to generate a posterior joint probability density function containing the relative contribution ratio of each source. Since this high-dimensional probability integral does not have an explicit analytical solution in the analytical domain, and the three or more endmembers of a single isotope constitute a mathematically underdetermined system, it is impossible to obtain a unique analytical solution through traditional algebraic equations.

[0074] The contribution rate estimation module 140 calls the Markov chain Monte Carlo algorithm to perform random sampling approximation in the parameter space. By performing random walks in the multidimensional probability space, it converges to calculate the maximum likelihood interval of the relative contribution ratio of each source under the joint posterior distribution, thereby solving the problem of quantitative analysis under multi-source mixture. For the underlying random walk and acceptance / rejection logic, it is based on the existing Gibbs sampling or Metropolis-Hastins algorithm.

[0075] To ensure that the sampling sequence truly converges to the target posterior distribution region and avoid algorithmic dead zones, the contribution rate estimation module 140 initiates at least three independent Markov chains in parallel and discards the first half of the total iteration steps as a burn period to eliminate the interference of the initial random state on the final result. During the iteration process, the contribution rate estimation module 140 calculates the potential scaling factor of each parameter in real time as a convergence diagnostic index. When this index meets the convergence judgment threshold condition, it indicates that the degree of difference between multiple parallel chains is basically the same as the degree of fluctuation within a single chain, and the model has reached a stable distribution. In this embodiment, the total number of iteration steps of a single Markov chain ranges from 100,000 to 500,000, preferably 300,000, to ensure that the parameter space is fully explored; the preset convergence judgment threshold ranges from less than or equal to 1.05.

[0076] ; In the formula: It is a potential scaling factor; The inter-sequence variance between parallel Markov chains; This represents the in-sequence variance within a single Markov chain.

[0077] When the iteration is completed and all diagnostic indicators of the relative contribution ratio parameters meet the convergence threshold conditions, the contribution rate estimation module 140 performs mathematical statistical analysis on the retained valid sample chain. Finally, the system extracts and outputs the posterior probability mean, median and corresponding confidence interval data of the relative contribution ratios of the three main anthropogenic emission sources: coal mining sources, natural gas system leakage and combustion sources, and coal chemical and industrial emission sources, thus completing the quantitative analysis of the source of atmospheric methane in the target city during winter.

[0078] To aid in understanding the technical solution of this invention, the following is an application example based on winter atmospheric observation data from a complex city in northern China.

[0079] The observation area was set as the central urban area of ​​the target city, and the time period was winter. The end-member library construction module 110 performed prior pruning based on the local winter average temperature and phenological conditions to confirm that methane emissions from natural wetlands and agricultural activities had been reduced to a negligible level. The end-member library finally identified three main anthropogenic emission sources and supplemented them by reviewing local historical measurement records and field measurements to determine their stable carbon isotope ratio parameters: the average value for coal mining sources was -33.88±2.31‰; the average value for natural gas system leakage and combustion sources was -25.62±1.45‰; and the average value for coal chemical and industrial emission sources was -47.06±0.41‰.

[0080] The sample acquisition and measurement module 120 performs timed sampling for several consecutive days within a specific time window from 14:00 to 14:15. After the instrument measures and removes the initial residual gas data from the pipeline, the average value of the signal during the stable period is extracted. The discrete measurement data of atmospheric methane concentration and stable carbon isotope ratio obtained during the observation period are recorded in the table below.

[0081] Table 1: Discrete Measurement Data of Atmospheric Methane Concentration and Stable Carbon Isotope Ratios during the Continuous Winter Observation Period The feature extraction module 130 uses the difference between the concentration in Table 1 and the background concentration (2.003 ppm) at a local global atmospheric station as the independent variable, and the product of the stable carbon isotope ratio sequence and the concentration in Table 1 as the dependent variable. The dataset is constructed by inputting the standard deviations of the independent and dependent variables. The module runs a bidirectional error regression algorithm to fit the model. As a control group, a traditional one-way least squares regression model is run simultaneously.

[0082] The contribution rate estimation module 140 imports the feature value parameters obtained from the regression into the Bayesian network model. The total number of iterations of the Markov chain Monte Carlo iterative algorithm is set to 300,000 for a single chain. The first 150,000 burn period sample points are discarded, and the posterior mean and 95% confidence interval of the relative contribution ratio of each source are extracted. To evaluate the effectiveness of the analytical method, the high-resolution emission inventory benchmark data officially released by the target area and the results calculated by traditional algebraic simultaneous equations are introduced as comparison objects. The comparison data are summarized in the table below.

[0083] Table 2: Comparison of Test Data Table of Mixed Source Characteristic Values ​​and Relative Contribution Ratios of Pollution Sources It should be noted that the high-resolution emission inventory baseline data in Table 2 was not generated by the bidirectional error regression and Bayesian network model of this embodiment, but was used as an external verification reference to evaluate the effectiveness of the inversion analysis of this scheme. The high-resolution emission inventory baseline data was obtained using the conventional bottom-up emission factor method in environmental statistics. The specific process is as follows: by collecting the baseline activity level data of the target area during the observation period, and combining it with the methane emission factor that is representative of the industry, and referring to the recommended factors such as the Tsinghua University MEIC multi-scale emission inventory model or the global EDGAR greenhouse gas emission database, the total absolute methane emission estimate of each single source was calculated, and then the relative contribution ratio of each mixed source was obtained by summing.

[0084] According to Tables 1 and 2, and Figure 5 It can be seen that the present invention's scheme improves the quantification accuracy of multi-source mixed systems and its correction effect compared to traditional algorithms. The discrete data distribution in Table 1 reflects the random noise of instruments and the fluctuation of environmental background in actual atmospheric physics monitoring. In the process of extracting the feature values ​​of newly added local mixed sources, the traditional one-way least squares method, due to unilaterally ignoring the methane concentration measurement error on the horizontal axis, causes a statistically attenuated bias in the slope of the fitted line, resulting in a drift of the calculated slope feature value and a large calculation variance. The two-way error regression model run by the feature value extraction module 130 of this scheme takes into account the measurement noise of both axes, suppresses systematic bias, and narrows the standard error of the feature value to 0.81‰. This data processing mechanism ensures the boundary determinism of the observation input end in the inversion system.

[0085] For the core task of multi-terminal analysis, a single isotope index forms an underdetermined system on the algebraic equations for three different emission sources. Table 2 shows that traditional algebraic methods cannot obtain a unique mathematical solution when dealing with unknown sources with more than one variable dimension, resulting in the overlap of industrial and natural gas sources and only outputting invalid lumped results. The contribution rate estimation module 140 of this scheme constructs a Bayesian probability mass balance control equation and calls the Markov chain Monte Carlo algorithm to perform random sampling integration of high-dimensional probability distributions, thus overcoming the solution limitations of the algebraic dimension and giving an independent source contribution probability distribution interval.

[0086] The "--" in Table 2 indicates that it is not applicable. Since the high-resolution emission inventory is a bottom-up statistical quantitative data, it is mainly used to provide a macro benchmark for the relative contribution ratio of each main controller to emission sources. It does not contain stable carbon isotope physical characteristic parameters of atmospheric environmental observation. Therefore, there is no corresponding comparative data in this item.

[0087] Comparing the posterior mean of the contribution ratio of each pollution source extracted by this scheme with the high-resolution emission inventory benchmark data reveals that although the two are consistent in terms of the dominant pollution type, there are significant dynamic differences in the specific values. For example, this scheme successfully quantified and captured the relatively high coal chemical and industrial fugitive emissions during the observation period, with a resolution value of 34.2%, which is significantly higher than the 10.7% of the static inventory. This difference precisely proves that the traditional static statistical inventory has a serious time lag, while the model of this invention can keenly reflect the real physical state of sudden leaks or dynamic emissions in the real atmospheric environment. Experimental test results show that by introducing environmental background concentration parameters to construct the Miller-Tans space, the system can separate weak signals of newly added pollution sources in complex urban monitoring environments. It also successfully overcomes the mathematical obstacles of algebraically underdetermined systems, and completes the quantitative decomposition and uncertainty assessment of the emission ratio of each pollution source under high-dimensional mixed conditions. This demonstrates the advanced nature, accuracy, and feasibility of the data processing logic of this technical solution in practical applications.

[0088] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for apportioning the sources of atmospheric methane in urban winters based on carbon isotopes, characterized in that, The method includes the following steps: Based on the target city's energy structure, industrial layout, policies and regulations, and meteorological conditions during winter, the main anthropogenic emission sources are identified, and a localized end-member database file containing the mean and standard deviation of stable carbon isotope characteristic values ​​of each main anthropogenic emission source is generated. Measure atmospheric samples and extract methane concentration sequences and stable carbon isotope ratio sequences; The methane concentration from a global atmospheric background reference station is obtained as the background concentration; A linear regression model was established with the difference between the methane concentration sequence and the background concentration as the independent variable and the product of the stable carbon isotope ratio sequence and the methane concentration sequence as the dependent variable, and the characteristic values ​​and standard errors of the newly added mixed source stable carbon isotopes were decoupled. The stable carbon isotope eigenvalues ​​and standard errors of the newly added mixed sources, along with the localized end-member database file, are loaded into the Bayesian mixed topology model to calculate and output the relative contribution ratio of each controlling human to the emission sources.

2. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 1, characterized in that, The steps for identifying the primary anthropogenic emission sources based on the target city's energy structure, industrial layout, policies, regulations, and meteorological conditions during winter include: Obtain the baseline emission flux of potential methane emission sources in the target city. If the contribution rate of a potential methane emission source to the baseline emission flux in the target season is lower than the preset contribution rate threshold, or if the biodegradation mechanism is significantly inhibited due to objective environmental factors, the potential source is removed. The sources of emissions are identified as refrigeration, heating, or industrial coal mining, natural gas system leaks and combustion sources, and coal chemical and industrial emission sources. The preset contribution rate threshold ranges from 1% to 5%.

3. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 2, characterized in that, The steps for eliminating the potential source when the biodegradation mechanism is significantly inhibited due to objective environmental factors include: When the winter temperature is lower than the preset temperature threshold, it is determined that the biological activity of methanogenic bacteria in wastewater treatment is significantly inhibited, thereby eliminating the waste treatment source. The temperature threshold ranges from 0°C to 4°C. Based on the seasonal characteristics of crop fallow, wetland freezing, and vegetation dormancy, if it is determined that the emissions from agricultural activity sources, wetland sources, and vegetation sources are significantly suppressed and the basic emission contribution rate is lower than the preset contribution rate threshold, then pruning and removal are performed.

4. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 1, characterized in that, The step of generating a localized end-member database file containing the mean and standard deviation of stable carbon isotope eigenvalues ​​for each master-controlled artificial emission source includes: Assuming that the stable carbon isotope ratio variable of any of the aforementioned main controllers as emission sources follows a normal distribution; when extracting measurement values ​​from historical literature data, if the sample size of the obtained data is lower than the preset sample size filtering limit, the on-site sampling device is controlled to collect gas samples for measurement; the preset sample size filtering limit is not less than 15.

5. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 1, characterized in that, Before the step of measuring the atmospheric sample, the constraints for obtaining the atmospheric sample include: The air intake is set at a height of 15 to 20 meters above the ground; and the daily physical sampling time window is set between 13:00 and 15:00 in the afternoon.

6. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 5, characterized in that, The atmospheric sample measurement steps are performed using a cavity ring-down spectrometer, and the measurement steps include: The continuous inlet gas flow rate of the optical cavity ring-down spectrometer is controlled to be 20 mL / min to 30 mL / min; the initial measurement data is discarded, and only the data from the stable period is extracted for mean calculation to extract the average methane concentration and the average stable carbon isotope ratio of the environmental atmospheric sample. After the observation period ends, the data are integrated into the methane concentration sequence and the stable carbon isotope ratio sequence; the initial measurement data rejection time is set to 5 to 10 minutes.

7. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 1, characterized in that, The steps for decoupling the eigenvalues ​​and standard errors of the newly added mixed source stable carbon isotopes include: A bivariate bidirectional error regression algorithm is used to fit the linear regression model to a straight line. During the iterative calculation process, hardware measurement errors in the horizontal and vertical directions are introduced. The measurement error weighting factors of each observation data point in the two axes are calculated respectively, and a comprehensive equivalent weighting coefficient containing bidirectional errors is constructed. The slope term of the fitted line calculated based on the optimal fit is used to decouple the characteristic values ​​of the newly added mixed source stable carbon isotopes.

8. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 7, characterized in that, After the step of decoupling the newly added mixed-source stable carbon isotope eigenvalues ​​and standard errors, the method further includes a step of performing regression quality filtering. The newly added mixed-source stable carbon isotope eigenvalues ​​are determined to be valid and the output is retained only if the following three-level constraint indicators are met simultaneously: Level 1 indicator: The concentration range threshold where the difference between the highest and lowest methane concentrations in the methane concentration sequence during the observation period is greater than 0.15 ppm to 0.35 ppm; Second-level indicator: The coefficient of determination of the optimal fitted line is greater than the threshold of 0.80 to 0.90; Level 3 metric: The standard error of the output slope is less than the upper limit threshold of the standard error of 1.5‰ to 2.5‰.

9. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 1, characterized in that, The steps of loading the newly added hybrid source stable carbon isotope eigenvalues, standard errors, and the localized endmember database file into the Bayesian hybrid topology model include: A probabilistic mass balance control equation is constructed, and the isotope fractionation coefficient and standard deviation of each of the main controlling human emission sources in the atmospheric transport process are set to zero; A non-information uniform prior distribution model is constructed using the Dirichlet distribution, wherein the concentration hyperparameter corresponding to each of the main controlling human emission sources in the Dirichlet distribution is set to 1.

10. The method for apportioning urban winter atmospheric methane sources based on carbon isotopes according to claim 9, characterized in that, The steps for calculating and outputting the relative contribution ratio of each main controller to the emission source include: The Markov chain Monte Carlo iterative algorithm program performs random sampling in the parameter space, starts at least three independent Markov chains in parallel, and calculates the potential scaling factor of each parameter in real time. When the potential scale reduction factor meets the convergence threshold condition, the determination model reaches a stable distribution. Mathematical statistics are performed on the retained effective sample chains, and the posterior probability mean of the relative contribution ratio of each of the main controllers to the emission sources is output.

Images